Time Series Forecasting
Chronos
Safetensors
t5
time series
forecasting
pretrained models
foundation models
time series foundation models
time-series
Instructions to use amazon/chronos-t5-mini with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Chronos
How to use amazon/chronos-t5-mini with Chronos:
pip install chronos-forecasting
import pandas as pd from chronos import BaseChronosPipeline pipeline = BaseChronosPipeline.from_pretrained("amazon/chronos-t5-mini", device_map="cuda") # Load historical data context_df = pd.read_csv("https://autogluon.s3.us-west-2.amazonaws.com/datasets/timeseries/misc/AirPassengers.csv") # Generate predictions pred_df = pipeline.predict_df( context_df, prediction_length=36, # Number of steps to forecast quantile_levels=[0.1, 0.5, 0.9], # Quantiles for probabilistic forecast id_column="item_id", # Column identifying different time series timestamp_column="Month", # Column with datetime information target="#Passengers", # Column(s) with time series values to predict ) - Notebooks
- Google Colab
- Kaggle
newlines
Browse files
README.md
CHANGED
|
@@ -19,7 +19,9 @@ For details on Chronos models, training data and procedures, and experimental re
|
|
| 19 |
|
| 20 |
## Architecture
|
| 21 |
|
| 22 |
-
The model in this repository is based on the [T5 architecture](https://arxiv.org/abs/1910.10683).
|
|
|
|
|
|
|
| 23 |
|
| 24 |
Model | Parameters | Based on
|
| 25 |
----------------|-------------------|----------------------
|
|
|
|
| 19 |
|
| 20 |
## Architecture
|
| 21 |
|
| 22 |
+
The model in this repository is based on the [T5 architecture](https://arxiv.org/abs/1910.10683).
|
| 23 |
+
The only difference is in the vocabulary size:
|
| 24 |
+
Chronos-T5 uses 4096 different tokens, compared to 32128 of the original T5 models, resulting in a smaller number of total parameters.
|
| 25 |
|
| 26 |
Model | Parameters | Based on
|
| 27 |
----------------|-------------------|----------------------
|